Add the posibility to type saved registers #284
Conversation
There was a problem hiding this comment.
Looks reasonable, although I think you might currently be ~the only user of --reg-vars so it's not going to see a lot of use. :) Extending it to setup_planned_vars or temp vars would make it a lot more useful, although, that looks a bit tricky, with the naming there being performed lazily...
| corresponding_regs = [ | ||
| r | ||
| for r in regs | ||
| if reg_field.name.endswith(str(r).removeprefix("$")) and not r.is_float() |
There was a problem hiding this comment.
reg_name = stack_info.function.reg_formatter.format(reg)
name == f"var_{reg_name}"I think. Why the and not r.is_float()?
Actually, hm, instead of hard-coding variable name, it might be cleaner if this function just parsed out a struct field name -> type mapping, and then add_register_var did the register -> struct field name lookup, since that is the one who knows the variable name.
| @@ -4365,6 +4368,34 @@ def setup_planned_vars( | |||
| stack_info.planned_vars[key] = var | |||
|
|
|||
|
|
|||
| def get_declared_reg_types( | |||
| stack_info: StackInfo, regs: List[Register] | |||
| ) -> [Dict[Register, Type]]: | |||
There was a problem hiding this comment.
this is invalid type syntax -- the [] need to be removed. Can you run mypy?
This is often useful when a void* is cast to a specific type and loaded into a saved register. It should be tested whether it causes issues in optimized code.